Monday, April 08, 2019

Section 512 roundtable part 3


SESSION 3: Domestic Developments
           
Richard James Burgess American Association of Independent Music: Not much different from 2016. Notice and takedown hasn’t helped with staydown; still need to send lots of notice. Optimistic about EU Copyright Directive.

Alex Feerst Medium: Runs Trust & Safety—moderators’ dilemma and the way that red flag knowledge hangs over a team trying to make a thoughtful and civil space while taking down violent/terrorist content.

Devlin Hartline Center for the Protection of Intellectual Property, Antonin Scalia Law School, George Mason University: 512 is not intended to be solely notice and takedown regime; incentives to cooperate to detect & deal with © infringement. Service providers were to play a role in preventing infringement, finding and removing it w/o input from © owners, through red flag knowledge standards. Lose safe harbors if aware of facts/circumstances from which infringing activity is apparent. Courts have read this so narrowly that a fire engine on fire would be necessary. Google’s ability to index the Pirate Bay is an example. Properly understood, red flag knowledge is general knowledge of infringement on the platform, not specific infringements, and the burden of investigating should fall on the service provider.

Catherine Gellis The Copia Institute: Don’t let the sky is falling rhetoric skew the recommendations. The sky is rising—the pie is growing, there is more revenue, more works due to the internet. Really careful not to kill the golden goose. Copia Institute just released a report aggregating a lot of the data providing the evidentiary record.  There’s still cause for alarm about pressures on platforms and individual creators: the effect of unadjudicated claims of infringement. This is the only area of law where reactions are required where there is never any adjudication.

Eric Goldman Santa Clara University School of Law: 512(f): pulled all of the cases since 2017; about 25 such cases, less than ½ substantively analyzed. I didn’t find any cases where a plaintiff actually won in court during the period; not inconsistent w/past jurisprudence, w/only 2 512(f) victories in 21 years.

Joseph Gratz Durie Tangri LLP: Not that many cases since the last set of roundtables. The system is basically working; the players have reached a rough modus vivendi leading to less large-scale litigation on the order of Giganews or Viacom.  The case law shows the flexibility of current law and the ability of courts to take into account specific situations, whether that is in the context of repeat infringers, expeditious removal, willful misrepresentation.

Jared Polin FroKnowsPhoto: YT channel/photographer.  I built everything off of free, hoping people will support me with what I sell.  I have two full time employees from the channel. 

Tamber Ray NTCA ― The Rural Broadband Association: 850 rural broadband providers, covering 33% of the US landmass, the sole provider in many of their communities. Hard to reconcile Cox/Grande and Peckingham cases. They have followed through on suspending repeat infringers, but are still getting the same number of infringement notices. Want guidance on responsibilities.

Robert Schwartz Consumer Technology Association: Toleration, accommodation, and implied license matter a lot. Overenforcement would hurt the creators it intended to help.

Christian Troncoso Business Software Alliance: Dual interests—content providers & hosts. There are frustrations on content protection side, but enterprise cloud service operators also have frustrations—tends to show that the compromise was fair in the end. End users are also important stakeholders subject to the outcomes here.

Kate Tummarello Engine: American ecosystem is thriving due to laws such as the DMCA. Startups support safe harbors to avoid ruinous infringement claims. Bad actors send false notices to startups and that’s a problem but we support the overall framework.

Ping Wang Freelancer/writer: user involvement.

Nancy Wolff   Digital Media Licensing Association: Fingerprinting tech is available, but little incentive to do any type of filtering, and little incentive to develop licensing systems. Curation makes a better website, but there aren’t licensing models.  [I have the sad feeling that most of the people here don’t know what Livejournal is/was; the description in the case comes close to describing the specific community at issue but has no relationship whatsoever to the journaling function that the site mainly and overall supports.]

Q re 512(f).

Gratz: not a sufficient deterrent for many kinds of abusive notices and counternotices. There haven’t been many litigated cases b/c what’s at stake isn’t always enough to support litigation. We may be seeing a greater number of those cases even under the current standard, not b/c of Lenz, but b/c of increasing and new ways that internet intermediaries are part of economic activity, as on Etsy. The competitive incentives to send bogus notifications and counternotifications will become so large that they’ll provide economic incentives for litigation.

Isbell: Etsy alluded to fraud in notices/counternotices; use for harassment.  Goldman mentioned several cases since we last heard from people but no plaintiffs have won.  Is that a good thing/bad thing?  Problem w/law or courts?

Goldman: data suggest that either there aren’t abuses worth litigating or that the law is miscalibrated to protect against those abuses, and the evidence suggests the latter. There’s no real incentive to do homework before sending a notice. There’s also often a lot of background dispute information, e.g., a dispute over a screenplay. The ISP is not in a good place to resolve that. The notice is a way for the parties to fight.

Greenberg: relatively few counternotices are filed; abusive counternotices exist too. 512(f) case of Johnson v. New Destiny Community Church—courts apply the same standards to both?

Goldman: correct. There are so many things that have to go right for a counternotice to be filed in the first place—the user has to know their rights and be willing to take the risk.  The system has little to encourage that.

Smith: if you’re focused on users being educated & confident enough, then would it make it harder for users to file counternotice if you loosen up standard for liability for false notices?

Goldman: they’re a nonfactor today so it wouldn’t have a material effect on the filing of counternotices. Focusing on abuses of notices is the place to start.

Polin: YT hits me with claims—many times I have permission or it’s fair use. It’s simple to file a counternotice on YT.  I generally win b/c we try to stick to good practices including on fair use.  YT does a good job for us.

Wolff: visual content online: the counternotice really puts the content owner at a disadvantage, particularly if the counternotice is in bad faith, b/c you can’t afford to go to court and file a claim w/in 10 days. If you don’t have a © registration before that, you have to spend $800 and find a lawyer. Improper counternotifications are harmful.

Wang: EU’s Directive provides an answer.  YT is only part of the ecosystem.  100s of my pieces are taken on other sites.  Google can fix it easily but they don’t because they earn money.

Q: where are these posted?

Wang: around the world where there are Chinese communities—different websites.  I don’t blame small forums w/a couple hundred people but Google has an easier solution.  I shouldn’t have to file a notice. They should have to ask permission before they use my stuff. 

Gratz: counternotices v. notices.  It’s not that anyone who fails to send a counternotice worries about 512(f). They worry about whether it will lead to litigation, even when they’re very certain they’d prevail. Even if you’re really sure that the dragon is wrong, you’re waking a sleeping dragon.

Gellis: Notice gaming is a problem—people are sending notices to ensure that speakers use up their strikes. The more important it is for a platform, esp. the larger platforms, to have a specific rubric about how many complaints they have to receive before they terminate a speaker, the easier it is to game that to eliminate a speaker that someone doesn’t like.

Greenberg: are there studies on this?

Gellis: anecdotal accounts.  [I collect these.  One link.  Another. There are also a bunch mentioned in Google’s additional comments from last time around.  EFF also submitted additional comments on this last time around, e.g.:
One OSP surveyed in the study by U.C. Berkeley School of Law researchers described “cases where allegedly abusive ex-husbands have filed DMCA complaints against images their ex-wife had posted as a means of attempting to get her current address.” According to an amicus brief filed on her behalf in Lenz v. Universal Music, when YouTube creator Rebecca Prince considered filing a counter-notice in response to an abusive notice, she “feared that gaining her sensitive information was in fact what this person wanted so he or she could use it to ‘dox’ her – that is, to release her identifying information online to further harass and intimidate her …

[RT continues: the most recent empirical study of which I am aware suggests that there are a lot of ungiven counternotices for plausibly noninfringing content, but that study doesn’t get to the level of sorting abuse v. mistake.  We also have a qualitative decision to make about how much we care about one instance of doxxing per X instances of accurate or wrong-but-reasonable notifications.]

Hartline: of course bogus takedowns are a problem, but hyperfocus on this misses the forest for the trees. Google’s transparency report: over 4 billion URLs, vast majority of which are illegit [which gets back to the basic point, which is that not everyone is Google—like Amazon Kindle, many sites receive more bogus takedowns than legit takedowns. If you make © policy for Google, you will ensure that only Google survives.]. Lenz was wrong b/c not having bad faith is not the same thing as having good faith.

Feerst: when we get a notice that seems defective/doesn’t seem to make sense, we often engage w/ the sender. There’s a large number of folks who don’t get the gravity of sending a takedown—they say things like “I’m not gonna do your job for you” and don’t provide a URL or a name.  We want to help folks give us information if content is truly bad. But the relative sloppiness of what we get is just a frustration.

Smith: it sounds like the “expeditious” flexibility you have is helping you figure out what’s really going on.

Feerst: the way you’re framing it, the risk that we have to take on is that “expeditious” might be defined later. We do take on additional risk to help folks.  Whether that’s fair is another Q.

Smith: you’re comfortable with the risk.

Feerst: not how he’d put it.

Isbell: automated notices v. forms: any change?

Burgess: concern about emphasis on hearsay and rumor.  Small content owners have given up.

Polis: as a small content creator, doesn’t want the gatekeepers shutting the gates again.  Some musicians may not be making as much money, but you have to use what’s here today. Lots of the musicians I photograph, I have a larger online presence than they do.

Q: re automated notices.

Ray: we get notices from multiple sources and can’t verify that they’re all legit. Or we get multiple notices for the same song for the same user and we can’t determine what is duplicative.  If we get multiple notices from different email accounts for the same user and same content, a P could say “you got 100 notices about this account on the same day and acted only on one.”  Huge recordkeeping burden.

Wolff: we’ve always had a level of human involvement—have to check whether a work was licensed or not.  She recommends that members say they considered fair use in their threat letters. The problem is the system itself; many have given up.

Greenberg: numbers of notices as strength v. weakness?

Gratz: not here for Google, though previously represented them. The number verges on meaningless b/c the vast bulk are for sites or items that were never indexed in the first place. People use the Google search result takedown form as a general “I haven’t seen this on Google, but this link infringes” form. So the number isn’t likely to be particularly meaningful especially at the difference between double-digit millions primarily sent by automated systems.

Troncoso: Ideal structure for takedown framework: part of the difficulty is the sheer diversity of stakeholders under the statute—search engines, social media, enterprise cloud companies that provide services to other 512(c) systems—coming up w/ a single system that works for all is different. So too with content, types of industries, types of licensing. That’s why we think the DMCA has done a fairly good job allowing for bespoke systems to arise. Google is doing things not required by the statute, like deindexing content before it appears. Google has the incentive to make sure they’re limiting the resources they have to throw at the problem before the links populate. Balancing the interests of all actors is not easy but all things considered the DMCA is pretty good.

Schwartz: Google is also a member of CTA, whose view has been that w/all these variables and difficulties the burden for initiation has to be on the rightsholder.

Goldman: a lot of activity is outside the scope of (c) and (f) altogether—the waning scope of both of those laws. First, the fastlane for rightsholders—more trusted interactions w/ the service and their actions aren’t covered by (c) or (f).  Second, prefiltering-type removals aren’t within those provision either.  Could we develop an optimal scheme for (c) or (f)?  First we have to recognize the decreasing percentage of these interactions actually covered.

Wong: Google claims to pay lots to creators and to remove links, but never tell us how much money they made by using content w/o permission.

Hartline: Lenz is wrong: absence of good faith isn’t bad faith. Wrong as a matter of policy: if you don’t have to consider fair use, why don’t you have to consider the other defenses? Even w/in Lenz, you can use an algorithm to assess whether you can take something down. It’s much harder to assess whether something is noninfringing than it is to assess whether it’s infringing.  [This is an obviously nonsensical statement that is justified by the larger explanation that fair uses are infringing-but-excused, but that’s not true even if you disregard the words of the statute (which you should not) and just think about the broader difference between a defense and excuse.  It is particularly interesting to find this argument made alongside the argument that the absence of good faith isn’t bad faith.  That one’s actually more understandable!]

Gratz: a good heart and an empty head is not good enough for 512.

Burgess: the DMCA is not working according to any content creators I know.  [The 1936 Literary Digest poll might have some lessons.]

Isbell: 512(h) and (j)?

Gellis: we’ve also ignored 512(h).  There hasn’t been a ton of caselaw percolating.  Not convinced there’s adequate protections built in for users, however we consider the subpoena interests. Due process right to anonymous speech should be meaningful.

Gratz: a large number of subpoenas from Strike 3 Holdings! In a number of other cases where subpoenas issued, many courts have because of potentially embarrassing nature of material at issue put protections in place to allow cases to proceed w/o permitting the public disclosure of the identity of the accused infringer. That’s an important development that helps separate out © from anonymity interest and protects against reputation threats used as leverage.

Wolff: porn cases skew decisions. Uploading content shouldn’t be equivalent to engaging in speech, if it’s not your own.  [I’ll remember that the next time I recite the kaddish.]  Can’t necessarily determine whether someone is even in the US.

Goldman: 512(h) is an artifact of a different time; wouldn’t be drafted the same way today. We don’t provide court-unsupervised access to others’ identity in other situations. Elsewhere, identifying a tortious actor requires a court’s permission. That’s not the deal we’d strike today; an anachronism. The fast lane has become a source of copyright trolling—lawsuits with the sole intent of extracting settlements. Noteworthy when a judge said that wouldn’t be allowed.  In the WHOIS context, the availability of info about domain name registrants won’t automatically be given, and most registrants use a privacy service. 2010s model would give more weight to privacy.

Isbell: 512(j): is that just DOA? Do we care? Should we care?

Goldman: Has logged every 512(j) case he’s seen; was supposed to be integral part of the structure. It was supposed to let ISPs avoid financial liability but allow © owners to get injunctive relief. They could do that and he doesn’t understand why that hasn’t been more widely explored.

Ray: 512(h): echoing previous statements—subpoenas are used not to protect © but to gain access to personal information used to threaten embarrassment. We want help figuring out what to do; some targets may be outside the court’s jurisdiction.

Gratz: puzzled about the same thing as Isbell.  Lack of 512(j) cases is an indication of how cooperation has happened between OSPs and takedown senders, not in the sense of holding hands and trying to do best possible solutions, but instead in that 512(j) allows you to force the OSP to do what it is otherwise incentivized to do—to take things down, terminate subscribers who are habitual infringers. Because those things happen anyway w/o the injunction, we don’t see a lot of those injunctions; their purpose has already been achieved.

Isbell: repeat infringers?

Tumarello: appreciate the flexibility of Motherless, but some certainty is really important for startups who want to know they’re doing the right thing, tell investors.

Isbell: what would do that? Statutory reform, best practices?

Tumarello: anything that was consensus-based and reflected the views of small companies. We’d want startups to have input.

Schwartz: We got into Cox as amicus on contributory infringement side. Generally, one thing that hasn’t been mentioned is the terrific abuse that was part of the notices that Cox received. They rejected the ones that had license offers attached to them; the courts have generally taken a dim view of those types of “notices.” That’s not to justify the other unfortunate things Cox did.

Gellis: Cox dealt with this head-on: Cox’s argument was that accruals for termination shouldn’t be contingent on anything but adjudicated claims. The problem w/ the system is that an allegation can have such an effect on a speaker and on a platform. The Cox court wasn’t worried but it is a big problem, because it turns into a prior restraint based on a nonadjudicated claim. Undercurrent: sometimes infringing is black and white, and sometimes noninfringing is, but often it’s not. Seeing a watermark, for example, is ambiguous.  You need more information in many cases! The senders are getting it wrong; if the senders do, the platforms have no hope of doing so either. They’re forced to defer to the complaint and presume it legitimate, which is hostile to expression and to the future rights of users—YouTube strikes accrued are serious concerns for speakers because their entire channel can be taken away.

Smith: Cox isn’t really about the expression of repeat infringers.  If we don’t want the perfect to be the enemy of the good, is there any room for these policies?

Gellis: as Schwartz says, there were significant problems with huge numbers of the claims Cox received and the district court rejected many of those notices.  But the court ignored the impact of having such a barrage of bad claims on Cox’s duty to regard the remainder of them as valid, despite its experience showing that so many were invalid in form and substance.  The court didn’t like Cox blacklisting Rightscorp but didn’t discuss why it had been blacklisted. Even if we want to treat filesharing differently, we can’t ignore the abuse there—and we can’t just transfer those rules to social media where there’s a lot more expression.

Gratz: yes, a continuum is critically necessary in the context of repeat infringer policies. Different kinds of ISPs, different kinds of subscribers, different kinds of activities—all factor into what kind of policy is reasonable and whether a particular user should be terminated. ISPs that care take that all into account, wanting to get rid of abusers but not noninfringing speech or users who make mistakes. The difference between conduit ISPs and edge services is very important here.  Consider the Copyright Alert system: a way of doing graduated response where the ultimate sanction was very serious: conduit termination. The reason Packingham is important is not necessarily b/c the First Amendment imposes limits on Congress’s ability to act here, though it may, but b/c it points out as a policy matter that these other things of importance to us as a society—child abuse—yield to the necessity of access to the internet.

Greenberg: (a) and (c) have different requirements in the statute, though Cox may change obligations of (a) services. How are conduits reconciling that their obligations may be not that different?

Gratz: From the POV of a conduit service—it’s still not a notice and takedown situation; it factors into whether someone is a repeat infringer.

Ray: Qs: what are our obligations to forward notices? Do we have to forward settlement demands? A big Q for our members.  Invalid DMCA notices, e.g. when they’re sent to our non-DMCA agent—can we disregard those when we have a valid DMCA agent & address?  Members are concerned w/Grande.  It put a pretty hefty obligation on ISPs w/o taking into account concerns about number of valid/invalid email addresses and notices. Also, what do we do with public wifi?  Tech helps some for P2P networking—availability of streaming services has changed member behavior—but what happens when members/subscribers have open wifi?  Subscribers are often happy to operate open wifi & another person can sit on the street and use it. What are we to do?  Also a need for coordination w/states/FCC.  Service is considered a fundamental service so disconnecting someone is a big deal to avoid trouble w/states.

Feerst: strategic uses of batching/attempts to silence users: we do see it in our Lumen notices. It’s not frequent but it’s real.  An investigative journalism publication out of Ecuador was using Flickr photos of politicians posted by the gov’t of Ecuador. Gov’t send takedown notices and a request to ban account for infringement.  That’s marginal now but suggests potential for mischief around repeat infringer policy.  The risk of having to go to summary judgment is a six-figure cost that can be deadly for a startup.  He had no guidance in making that risk determination.

No comments: